PREDICTION OF LITHUANIAN GDP: ARE REGRESSION MODELS OR TIME SERIES MODELS BETTER?

Abstract

The problem of Lithuanian GDP prediction is relevant. There are several institutions, such as Statistics Lithuania, state’s central bank, other banks that constantly announce their predictions of GDP. Frequently the forecasts of different institutions vary because they use different methods. The main purpose of the paper is to investigate whether regression models made of the monthly published economic indicators or time series models are better for Lithuanian GDP prediction. The changes of Lithuanian GDP as well as many other economic indicators that have impact on GDP are published quarterly. Prediction of quarterly economic indicators as it was done by the most researchers can be related to greater errors comparing with the prediction models that are made according to the monthly data. Monthly data can ensure that the newest information is used for prediction of GDP and show how the state’s economy is changing in the current quarter, that’s why it can reduce the error of prediction. The research is based on the economic data that is measured and published monthly by Statistics Lithuania (154 ratios at all). Various linear and non-linear regression models are made in order to find the best model for Lithuanian GDP prediction. The results of regression models are also compared with the results got by ARIMA (time series) models. The analysis showed that the regression models made of monthly published economic indicators may be better than time series models for prediction or Lithuanian GDP. DOI: http://dx.doi.org/10.5755/j01.em.18.4.5041